Modern companies are actively usingInternet applications for identifying users by face, which allows them to form huge databases and actually violate privacy. To counteract such systems based on neural networks, an application created at the University of Chicago will be able to avoid identification by photograph and disrupt the work of algorithms of programs like Clearview AI.
User image when appliedFawkes applications mask a photo at the pixel level, replacing certain points in the image with fragments borrowed from other photos. Thus, for example, a fragment of a photograph is changed to an element taken from an image that is as unlike the original as possible and even from a "donor" of a different gender. Changes in the photo are almost invisible to users, but they disrupt the work of the face search algorithm.
According to the developers, the useFawkes' applications have already made it possible to "trick" the search engines of Amazon, Microsoft and the Chinese Megvii. Further development of the application will avoid identification by the Clearview AI algorithm, already used by the police and private companies for personal identification based on a multi-billion dollar database.
University of Chicago Professor Ben Zhaostated that Fawkes will "make Clearview go" in the future. However, the Fawkes app is currently under development and does not always perform well. Further improvement of the program as conceived by the developers may lead to the fact that users will always process photos before uploading them to the network. Also, the integration of the application into major social networks will prevent applications like Clearview from conducting unauthorized identification of the person from user photos.